# A Characterization of Commutative Clean Rings

```A Characterization of Commutative Clean Rings
Warren Wm. McGovern
Department of Mathematics and Statistics
Bowling Green State University
Bowling Green, OH 43403
USA
[email protected]
Abstract.
A commutative ring A is said to be clean if every element of A can be
written as a sum of a unit and an idempotent. This definition dates back to 1977 where
it was introduced by W. K. Nicholson [7]. In 2002, V. P. Camillo and D. D. Anderson [1]
investigated commutative clean rings and obtained several important results. In [4] Han
and Nicholson show that if A is a semiperfect ring, then A[Z2 ] is a clean ring. In this paper
we generalize this argument (for commutative rings) and show that A[Z2 ] is clean if and
only A is clean. We also show that if the group ring A[G] is a commutative clean ring, then
G must be a torsion group. Our investigations lead us to introduce the class of 2-clean
rings.
Keywords: commutative clean ring, commutative group ring
2
Warren Wm. McGovern
1. 2-Clean Rings
Definition 1.1. The element a ∈ A is said to be clean if there exists an idempotent e ∈ A
such that a − e is invertible. If every element of A is clean, then A is said to be a clean ring.
The set of units and idempotents of A shall be denoted by U(A) and Id(A), respectively. As
usual M ax(A) denotes the set of maximal ideals of A, and it is equipped with the hull-kernel
topology. This means that the collection of sets of the form V (a) = {M ∈ M ax(A) : a ∈ M }
for a ∈ A form a base for the closed sets. The complement of V (a) is denoted by U (a).
It is well known that von Neumann regular rings, and more generally zero-dimensional rings
are clean. Also, local rings are clean, and, in fact, are precisely the indecomposable clean rings.
Thus, if A is an integral domain, then A is clean if and only if A is local. Some fundamental
facts regarding clean rings are that every homomorphic image of a clean ring is a clean ring,
and that a ring A is clean if and only if A/n(A) is clean where n(A) denotes the nilradical of
A. Also, a direct product of rings is clean if and only if each factor is clean. Most of these
facts can be found in [4] or [1]. We urge the interested reader to check there for their proofs.
All rings are commutative and with identity.
In [6] a list of several characterizations of commutative clean rings is given. Included in
the list is one given by Johnstone [5] which we presently state.
Theorem 1.2 [Johnstone]. A is a clean ring if and only if M ax(A) is zero-dimensional and
every prime ideal is contained in a unique maximal ideal.
A topological space is said to be zero-dimensional if it has a base of clopen sets. It is
known that if e ∈ Id(A), then U (e) is a clopen subset of M ax(A) but the reverse is not true
in general. A clopen subset of M ax(A) of the form U (e) for some e ∈ Id(A) will be called an
idempotent clopen. In [6] it is shown that M ax(A) is zero-dimensional and every clopen is an
idempotent clopen precisely when A is a clean ring.
Theorem 1.3. The following are equivalent for a ring A
(i) A is a clean ring.
(ii) For each a ∈ A there exists an idempotent e such that V (a) ⊆ U (e) and V (a − 1) ⊆ V (e).
(iii) The collection of idempotent clopen subsets of M ax(A) is base for the topology on
M ax(A).
(iv) M ax(A) is zero-dimensional and every clopen subset is of the form U (e) for some e ∈
Id(A).
Here is another characterization of a clean ring.
Proposition 1.4. A is clean if and only if for every a ∈ A, there is an idempotent e ∈ A such
that both a + e and a − e are invertible.
Proof. The sufficiency is clear. As for the necessity, let a ∈ A. By Theorem 1.3, choose an
idempotent e ∈ A such that V (a2 ) ⊆ U (e) and V (a2 − 1) ⊆ V (e). Let M be any maximal
ideal of A. If a ∈ M , then so is a2 and hence e ∈
/ M . It follows that a + e, a − e ∈
/ M . If
a, e ∈
/ M , then neither a + 1 nor a − 1 belong to M . Thus, a + e + M = a + 1 + M 6= M and
a − e + M = a − 1 + M 6= M , whence a + e, a − e are invertible.
A Characterization of Commutative Clean Rings
3
Definition 1.5. We call a commutative ring A 2-clean if for every pair of elements a, b ∈ A
there exists an idempotent e ∈ A such that a + b − e and a − b − e are both invertible. Observe
that letting b = 0 we obtain that a 2-clean ring is clean. Also, if charA = 2, then A is clean if
and only if A is 2-clean.
The proof of our next proposition is straightforward.
Proposition 1.6. 1) A homomorphic image of a 2-clean ring is 2-clean.
2) A direct product A = Πi∈I Ai is 2-clean if and only if each factor Ai is 2-clean.
Proposition 1.7. The following are equivalent for a ring A.
(i) A is 2-clean.
(ii) A/n(A) is 2-clean.
(iii) A is clean and A/J(A) is 2-clean.
(iv) A/J(A) is 2-clean and idempotents lift modulo the Jacobson radical.
Proof. The proof is exactly the same as the one used to show that a ring A is clean if and
only if A/n(A) is clean. We leave the verification to the interested reader. Our next result gives an easier way of checking whether a clean ring is 2-clean.
Proposition 1.8. A is 2-clean if and only if A is clean and 2 ∈ J(A).
Proof. Necessity. If A is 2-clean, then it is clean. Let a ∈ A be an arbitrary element. Using
the 2-clean condition on the pair a, a there is an idempotent e ∈ A such that 2a − e and −e
(and hence e) are invertible. The only idempotent that is invertible is e = 1. Therefore, 2a − 1
is invertible, whence 2 ∈ J(A).
Sufficiency. Let a, b ∈ A. Choose an idempotent e ∈ A such that a + b − e is invertible. We
claim that a − b − e is also invertible. Let M be a maximal ideal of A. By hypothesis, 2 ∈ M
so that a − b − e + M = a + b − e + M 6= M . Therefore, a − b − e ∈
/ M for any maximal ideal
of A. It follows that both a − b − e and a + b − e are invertible.
Theorem 1.9. Let A be a zero-dimensional ring. Then the following are equivalent:
(i) A is 2-clean.
(ii) charA = 2k for some natural k.
(iii) 2 is nilpotent.
(iv) For each maximal ideal M of A, charA/M = 2.
Proposition 1.10. Suppose A is a local ring. Then A is 2-clean if and only if charA/M = 2.
Proof. The necessity is clear so suppose that charA/M = 2. This means that a + b + M =
a − b + M for all a, b ∈ A. Therefore, a + b ∈ M if and only if a − b ∈ M . Thus, if both
a + b, a − b ∈ M , then a + b − 1, a − b − 1 ∈
/ M and hence both are invertible. Otherwise,
a + b, a − b ∈
/ M means that they are both invertible.
4
Warren Wm. McGovern
2. Commutative Clean Group Rings
In this section we address the question of when a group ring is clean. Some work on this
question was done in [4] where they showed that if A is a boolean ring and G is a torsion
group, then A[G] is a clean ring. We will only consider the case where the group in question is
abelian. When a group ring is zero-dimensional or local has been characterized and therefore
we have some sufficient conditions for a group ring to be a clean ring. We list [3] as our main
reference on commutative group rings, though we warn the casual reader that most of the
information in [3] is done for semi-group rings.
All groups are abelian.
Definition 2.1. Let A be a ring and (G, +) a group. The group ring is the set of formal sums
of the form Σni=1 ai X gi under pointwise addition and multiplication defined by X g X h = X g+h
on monomials and extending this in the obvious way.
Theorem 2.2. A[G] is zero-dimensional if and only if A is zero-dimensional and G is a torsion
group.
For commutative rings and abelian groups we obtain the following corollary.
Corollary 2.3. If A is a boolean ring and G is a torsion group, then A[G] is a clean ring.
Theorem 2.4. A[G] is local if and only if A is local, charA/M = p, and G is a p-group. In
particular, if A[G] is local, then G is torsion.
Our aim is to show that for A[G] to be a clean ring it is necessary that A be clean and G
be torsion. The first part is clear as A is a homomorphic image of A[G]. To obtain the second
part we need a series of lemmas. The first is well known.
Lemma 2.5. A[G] is an integral domain if and only if A is an integral domain and G is torsion
free.
Lemma 2.6. For any ring A and group G, A[G] is clean if and only if A/n(A)[G] is clean.
Proof. The necessity is obvious since A/n(A)[G] is a homomorphic image of A[G]. As for
the sufficiency it is known that n(A)[G] ≤ n(A[G]), which means that A[G]/n(A[G]) is a
homomorphic image of A/n(A)[G] and therefore A[G] is clean.
Proposition 2.7. If A[G] is a clean ring, then A is clean and G is a torsion group.
Proof. Clearly, A is a clean ring. Let t(G) denote the torsion subgroup of G. Let P be a
prime ideal of A. Then (A/P )[G/t(G)] is a homomorphic image of A[G] and hence is clean.
Since G/t(G) is a torsion-free group and A/P is an integral domain it follows by the Lemma 2.5
that (A/P )[G/t(G)] is a clean integral domain and hence is local. By Theorem 2.4 it follows
that G/t(G) is torsion, whence G = t(G).
Remark 2.8. The converse of the previous proposition is certainly not true. In [4] it is shown
that if A is the localization of Z at the prime 7Z (A is clean because it is local) and G = Z3
(clearly torsion), then A[G] is not clean. In general, discovering when A[G] is clean has eluded
us except for in certain cases. In the next section we consider the specific case when G = Z2 .
A Characterization of Commutative Clean Rings
5
Proposition 2.9. Suppose A is zero-dimensional. Then the following are equivalent:
(i) A[G] is clean.
(ii) A[G] is zero-dimensional.
(iii) G is a torsion group.
Proof. This is in fact a corollary to Proposition 2.7 and Theorem 2.2.
Corollary 2.10. Let F be a field. Then F [G] is clean if and only if G is torsion.
3. When G = Z2
In this section we investigate the cleanliness of the group ring A[Z2 ].
Definition 3.1. Let e be an idempotent of A. The element eX 0 ∈ A[G] is an idempotent
of A[G]. Any idempotent of A[G] of this form is said to be an idempotent in A. If every
idempotent of A[G] is in A, then we say that the idempotents of A[G] are in A.
Lemma 3.2. The element rX 0 + sX 1 ∈ A[Z2 ] is invertible if and only if r + s, r − s ∈ U(A).
Proof. Suppose rX 0 +sX 1 ∈ U(A[Z2 ]). Then there are f, g ∈ A such that (rX 0 +sX 1 )(f X 0 +
gX 1 ) = 1. This means that rf + sg = 1 and rg + sf = 0. Adding these two equations gives us
that (r + s)(f + g) =. Subtracting yields (r − s)(f − g) = 1. Therefore, r + s, r − s ∈ U(A).
r
−a
Conversely, if r + s, r − s ∈ U(A), then so is r2 − s2 . Let f = r2 −s
2 and g = r 2 −s2 . A quick
check shows that rf + sg = 1 and rg + sf = 0. It follows that rX 0 + sX 1 is an invertible
element of A[Z2 ].
Lemma 3.3. The element cX 0 + dX 1 ∈ A[Z2 ] is an idempotent if and only if c2 + d2 = c and
2cd = d. In this case, c + d, c − d ∈ Id(A).
Proof. The first statement follows from the equation cX 0 + dX 1 = (cX 0 + dX 1 )2 = (c2 +
d2 )X 0 + 2cdX 1 . The second is obtained by adding (and subtracting) the equations.
Proposition 3.4. The following are equivalent for the ring A:
(i) A is clean.
(ii) For every a ∈ A, the element aX 0 ∈ A[Z2 ] can be written as the sum of a unit and an
idempotent in A.
(iii) For every a ∈ A, the element aX 1 ∈ A[Z2 ] can be written as the sum of a unit and an
idempotent in A.
Proof. (i) ⇒ (ii) If A is clean, then for each a ∈ A let u ∈ U(A) and e ∈ Id(A) such that
a = u + e. Then
aX 0 = uX 0 + eX 0
is a representation of aX 0 as a sum of a unit of and an idempotent in A.
6
Warren Wm. McGovern
(ii) ⇒ (i) Let a ∈ A. If
aX 0 = (rX 0 + sX 1 ) + eX 0
is a representation of aX 0 as a sum of a unit and an idempotent in A, then s = 0 and thus
r = r + s ∈ U(A). Therefore, A is a clean ring.
(iii) ⇒ (i) Let a ∈ A. Suppose that aX 1 is the sum of a unit and an idempotent in A, say
aX 1 = (rX 0 + aX 1 ) + eX 0 .
for some e ∈ Id(A). By Lemma 3.2 r + a, a − r ∈ U(A). Since r = −e it follows that a − e and
a + e are both invertible and so by Proposition 1.4 A is clean.
(i) ⇒ (iii) Let a ∈ A. By Proposition 1.4 there exists an e ∈ Id(A) such that a − e, a + e
are invertible. Set r = −e so that
aX 1 = (rX 0 + aX 1 ) + eX 0 .
Since r + a = a − e and r − a = −(a + e) are invertible it follows that aX 1 is the sum of a unit
and an idempotent in A.
Proposition 3.5. For every a ∈ A, the element aX 0 + aX 1 ∈ A[Z2 ] can be written as the
sum of a unit and an idempotent in A if and only if 2 ∈ J(A). In this case every idempotent
of A[Z2 ] belongs to A.
Proof. Suppose aX 0 + aX 1 ∈ A[Z2 ] can be written as the sum of a unit and an idempotent
in A for every a ∈ A, say
ax0 + aX 1 = (rX 0 + sX 1 ) + eX 0 .
Since s = a it follows that r + a, r − a ∈ U(A) by Lemma 3.2. Notice that e = a − r which is
invertible and hence e = 1 and r = a − 1. Thus r + a = 2a − 1 ∈ U(A) for every a ∈ A, whence
2 ∈ J(A).
Conversely, let a ∈ A and 2 ∈ J(A). Since
aX 0 + aX 1 = ((a − 1)X 0 + aX 1 ) + X 0
and (a − 1) + a = 2a − 1 and (a − 1) − a = −1 are both invertible, the result follows from
Lemma 3.2.
Finally, suppose 2 ∈ J(A) and let aX 0 + bX 1 be idempotent. This means that a2 + b2 = a
and 2ab = b. The latter equality implies that (2a − 1)b = 0. By hypothesis, 2 ∈ J(A) and
therefore 2a − 1 is invertible. It follows that b = 0 and so the idempotents of A[Z2 ] belong to
A.
Proposition 3.6. The following are equivalent.
(a) For each a ∈ A, the element aX 0 + aX 1 is clean.
(b) For each a ∈ A there exists an e ∈ Id(A) and t, u ∈ A such that (2a − 1)t = (1 − e), and
2au = e.
(c) For each a ∈ A, there exists an e ∈ Id(A) such that V (2a) ⊆ V (e) and V (2a − 1) ⊆ U (e).
A Characterization of Commutative Clean Rings
7
Proof. (b)⇒(a). Let a ∈ A and choose e ∈ Id(A) and t, u ∈ A satisfying (b). Let f = 1 − e
so that ef = 0 and e + f = 1. Observe that since both e and f are idempotent we may without
loss of generality assume that ue = u, and tf = t. Let x = aue and observe that 2x = e and
ex = x. Let y = (a − 1)f , z = (a − x)e, and r = y + z. Next let d = −x and c = d + 1 = 1 − x.
Now,
c2 + d2 = d2 + 2d + 1 + d2 = 2d2 + 2d + 1 = 2x2 − 2x + 1 = x + f.
Notice that 2x = e = 1 − f so that c = 1 − x = x + f and therefore c2 + d2 = c. Next,
2cd = 2c(−x) = −ce = −(1 − x)e = ex − e = x − e = x − 2x = −x = d
and so cX 0 + dX 1 is an idempotent. Now,
r+c = y+z+c
= (a − 1)f + (a − x)e + 1 − x
= af − f + ae − ex + 1 − x
= a − f − ex + 1 − x
= a + e − 2x
= a
so that
aX 0 + aX 1 = (rX 0 + (r + 1)X 1 ) + (cX 0 + dX 1 ).
What is left to be shown is that (rX 0 + (r + 1)X 1 ) is invertible. To that end, first notice that
2r + 1 = 2y + 2z + 1
= 2(a − 1)f + 2(a − x)e + e + f
= (2a − 1)f + 2ae.
Next,
(2r + 1)(t + u) = ((2a − 1)f + 2ae)(t + u)
= (2a − 1)t + 2aue
= f +e
= 1.
Therefore rX 0 + (r + 1)X 1 is invertible by Lemma 3.2, whence aX 0 + aX 1 is clean.
(a)⇒(c). Suppose aX 0 + aX 1 is clean and write it as a sum of a unit and idempotent, say
aX 0 + aX 1 = (rX 0 + sX 1 ) + (cX 0 + dX 1 ).
Note that a = r+c = s+d. Since (rX 0 +sX 1 ) is invertible, r+s ∈ U(A), and since (cX 0 +dX 1 )
is an idempotent, it follows that c+d ∈ Id(A). Now 2a = a+a = r+c+s+d = (r+s)+(c+d) and
so 2a is a clean element. It follows that there exists an idempotent e such that V (2a) ⊆ V (e)
and V (2a − 1) ⊆ U (e).
The equivalence of (b) and (c) is straighforward.
Theorem 3.7. For a ring A, the following are equivalent:
8
Warren Wm. McGovern
(i) A is 2-clean.
(ii) A[Z2 ] is clean and the idempotents of A[Z2 ] belong to A.
(iii) A[Z2 ] is clean and 2 ∈ J(A).
(iv) Every element of A[Z2 ] can be written as the sum of a unit and an idempotent in A.
(v) A is clean and for every a, b ∈ A, V (a + b) ∩ V (a − b − 1) = ∅.
Proof. (i)⇒ (ii) If A is 2-clean, then by Proposition 1.8 and Proposition 3.5 every idempotent
of A[Z2 ] is in A. Let aX 0 + bX 1 ∈ A[G]. By the hypothesis there is an idempotent e such that
a + b − e and a − b − e are both invertible. Set r = a − e so that r + b and r − b are invertible.
It follows that
aX 0 + bX 1 = (rX 0 + bX 1 ) + eX 0 .
is a clean representation,whence A[Z2 ] is a clean ring.
(ii) ⇒ (iii) Suppose that A[Z2 ] is clean and that idempotent elements are in A. Let a ∈ A
and write
aX 0 + aX 1 = (rX 0 + aX 1 ) + eX 0
where e2 = e, r + e = a, and rX 0 + aX 0 is invertible. As previously noted it follows that
r + a, r − a are both invertible. Since r − a = −e is invertible it follows that e = 1. Therefore,
the invertibility of 2a − 1 = r + a and arbitrariness of a imply that 2 ∈ J(A).
(iii) ⇒ (ii) This is patent.
(ii) ⇒ (i) Let a, b ∈ A. Then the element aX 0 + bX 1 is clean and thus we can write it as
aX 0 + bX 1 = (rX 0 + bX 1 ) + eX 0
where e ∈ Id(A) and (rX 0 + bX 1 ) ∈ U(A[Z2 ]). Observe that r + e = a and r + b, r − b ∈ U(A).
Since
a + b − e = r + b and a − b − e = r − b
we are done.
(ii) ⇒ (iv). This is clear.
(iv) ⇒ (i). Let a, b ∈ A. By hypothesis there is a unit rX 0 + sX 1 ∈ A[Z2 ] and an
idempotent e ∈ A such that
aX 0 + bX 1 = (rX 0 + sX 1 ) + eX 0 .
It follows that s = b and that a − e = r. It is straighforward from here to show that A is
2-clean.
(i) ⇒ (v). If A is 2-clean, then V (a − b − 1) = V (a + b − 1) which is always disjoint from
V (a + b).
(v) ⇒ (1). For any a ∈ A we have ∅ = V (a−a)∩V (a+a−1) = V (0)∩V (2a−1) = V (2a−1)
and therefore 2a − 1 is invertible. It follows that 2 ∈ J(A).
Corollary 3.8. Let charA = 2. Then A[Z2 ] is clean if and only if A is clean.
A Characterization of Commutative Clean Rings
9
The next result was shown in [4]. We include a sketch of proof for completeness sake.
Proposition 3.9. Let 2 ∈ U(A). Then A[Z2 ] ∼
= A × A. In this case A is clean if and only if
A[G] is clean.
Proof. Define ψ : A[G] → A × A by ψ(aX 0 + bX 1 ) = (a + b, a − b). Then ψ is a ring
homomorphism. When 2 is a regular element this homomorphism is injective, and when 2 is
invertible the map is a surjection.
Remark 3.10. If V (2) is a clopen subset of M ax(A), then there is a decomposition of A =
A1 ⊕ A2 where the rings A1 and A2 satisfy that 2 ∈ U(A1 ) and 2 ∈ J(A2 ). It follows that A[Z2 ]
is a clean ring, but not every element of A[Z2 ] is the sum of a unit and a normal idemptotent.
Lemma 3.11. Suppose cX 0 + dX 1 ∈ Id(A[Z2 ]). Then both V (c), V (d) r V (c) are idempotent
clopen subsets of M ax(A). Moreover, V (d) is an idempotent clopen subset of M ax(A).
Proof. The second statement clearly follows from the first. By hypothesis we know that
c2 + d2 = c, 2cd = d, and c + d, c − d ∈ Id(A). The second equation implies that V (2), V (c) ⊆
V (d). In particular, we obtain that V (c) ⊆ V (c+d)∩V (c−d). Now, if M ∈ V (c+d)∩V (c−d),
then both 2c ∈ M and c2 − d2 ∈ M . If 2 ∈ M , then the residue field has characteristic 2 and
so
c + M = c2 + d2 + M = c2 − d2 + M = M
so that c ∈ M . It follows that V (c) = V (c + d) ∩ V (c − d), and so V (c) is an idempotent clopen
subset of M ax(A).
Next we show that V (c+d)∪V (c−d) = U (d)∪V (c) from which it will follow that V (d)rV (c)
is an idempotent clopen subset of M ax(A). Suppose that d ∈
/ M . Then 2c + M = 1 + M .
From this and c2 + d2 = c we conclude that 2d + M = 1 + M or 2d + M = −1 + M . (Observe
2
2
that in any field of characteristic different from 2 we have that 21 − 12 = 12 ). Therefore,
c + M = d + M or c + M = −d + M in which case M ∈ V (c − d) or M ∈ V (c + d), whence
U (d) ∪ V (c) ⊆ V (c + d) ∩ V (c − d). The reverse is the easy case.
Theorem 3.12. For a ring A, A[Z2 ] is clean if and only if for each a, b ∈ A there exist
idempotents e1 , e2 , e3 , e4 ∈ Id(A) such that U (ei ) ∩ U (ej ) = ∅, i 6= j,
1 = e1 + e2 + e3 + e4
(1)
V (2) ⊆ U (e3 ) ∪ U (e4 )
(2)
V (a + b) ⊆ U (e3 ) ∪ U (e1 )
(3)
V (a + b − 1) ⊆ U (e4 ) ∪ U (e2 )
(4)
V (a − b) ⊆ U (e3 ) ∪ U (e2 )
(5)
V (a − b − 1) ⊆ U (e4 ) ∪ U (e1 )
(6)
Proof. Necessity. Suppose A[Z2 ] is clean and let a, b ∈ A. Then we can write
aX 0 + bX 1 = (rX 0 + sX 1 ) + (cX 0 + dX 1 )
where rX 0 + sX 1 ∈ U (A[Z2 ]) and (cX 0 + dX 1 ) ∈ Id(A[Z2 ]). This means that c2 + d2 = c,
2cd = d, c + d, c − d ∈ Id(A) and both r + s and r − s are invertible. By Lemma 3.11 it follows
10
Warren Wm. McGovern
that there are idempotents e1 , e2 , e3 , e4 such that U (e4 ) = V (c), U (e3 ) = V (d) r V (c), U (e1 ) =
V (c − d) r
S V (c) and U (e2 ) = V (c + d) r V (c). Observe that U (ei ) ∩ U (ej ) = ∅ for each i 6= j
and that 1≤i≤4 U (ei ) = M ax(A). Without loss of generality we can assume that ei ej = 0 so
that e1 + e2 + e3 + e4 is an invertible idempotent and hence (1) is satisfied. That (2) is satisfied
follows from the fact that V (2) ⊆ V (d) = U (e3 ) ∪ U (e4 ).
Next, notice that a + b − (c + d) = r + s and so (3) V (a + b) ⊆ U (c + d) = U (e1 ) ∪ U (e3 ).
Similarly, a − b − (c − d) = r − s and so (5) V (a − b) ⊆ U (c − d) = U (e2 ) ∪ U (e3 ). Let
M ∈ V (a + b − 1). If M ∈ U (e1 ) = V (c − d) r V (c), then c + d ∈
/ M (otherwise 2c ∈ M and
2∈
/ M would imply c ∈ M .) Since c + d is idempotent, c + d + M = 1 + M . Hence
M = a + b − 1 + M = a + b − (c + d) + M
contradicting that r + s is invertible. Similarly, if M ∈ U (e3 ) = V (d) r V (c) then d + M = M
and so c + M = 1 + M . Therefore
M = a + b − 1 + M = a + b − (c + d) + M.
Therefore, (4) is satisfied. A similar argument yields (6).
Sufficiency. Let aX 0 +bX 1 ∈ A[Z2 ] and suppose that e1 , e2 , e3 , e4 are idempotents satisfying
the hypothesis. Observe that ei ej = 0 for all i 6= j. We shall produce elements r, s, c, d ∈ A
such that c2 + d2 = c, 2cd = d, and both r + s = a + b − (c + d) and r − s = a − b − (c − d) are
invertible. Thus, it follows that
aX 0 + bX 1 = (rX 0 + sX 1 ) + (cX 0 + dX 1 )
is a clean representation. To that end observe that (1) implies there is an x ∈ (e1 + e2 )A such
that 2x = e1 + e2 . We further suppose that x(e1 + e2 ) = x. Let y1 = xe1 and y2 = −xe2 . Set
c = x + e3 and d = y1 + y2 . Observe that
V (c) = U (e4 ) and V (d) = U (e4 ) ∪ U (e3 ).
Since
c2 + d2 = (x + e3 )2 + (y1 + y2 )2
= x2 + e3 + x2 e1 + x2 e2
= 2x2 + e3
= x + e3
= c
and
2cd = 2(x + e3 )(xe1 − xe2 )
= 2(x2 e1 − x2 e2 )
= 2x2 (e1 − e2 )
= x(e1 + e2 )(e1 − e2 )
= x(e1 − e2 )
= d
A Characterization of Commutative Clean Rings
11
it follows that cX 0 + dX 1 ∈ Id(A[Z2 ]). By construction c + d = 2xe1 + e3 so that
V (c + d) = U (e2 ) ∪ U (e4 ).
Let r = a − c and s = b − d. Then r + s = a + b − (c + d) and r − s = a + b − (c − d).
If r + s is not invertible, then it lies in some maximal ideal M . If M ∈ V (a + b), then
M ∈ V (c + d) contradicting (3). Therefore, M ∈ U (a + b) and hence M ∈ U (c + d). So
a + b + M = c + d + M = 1 + M , i.e. M ∈ V (a + b − 1). By (4) M ∈ U (e2 ) (since
M∈
/ V (c) = U (e4 )). Our contradiction is
1 + M = c + d + M = xe1 + xe2 + e3 + xe1 − xe2 + M = 2xe1 + xe3 + M = M
since M ∈ V (e1 ) ∩ V (e3 ). Therefore, r + s is invertible. A similar argument shows that r − s
is invertible.
Theorem 3.13. A[Z2 ] is clean if and only if A is clean.
Proof. As we have pointed so many times already, if A[Z2 ] is clean, then so is A. So
suppose that A is clean and let a, b ∈ A. Choose idempotents f1 , f2 ∈ A such that V (a + b) ⊆
U (f1 ), V (a + b − 1) ⊆ V (f1 ) and V (a − b) ⊆ U (f2 ), V (a − b − 1) ⊆ V (f2 ). Let e1 , e2 , e3 , e4 be
idempotents such that
U (e1 ) = U (f1 ) ∩ V (f2 ),
U (e2 ) = U (f2 ) ∩ V (f1 ),
U (e3 ) = U (f1 ) ∩ U (f2 ),
U (e4 ) = M ax(A) r (U (e1 ) ∪ U (e2 ) ∪ U (e3 )).
Observe that U (ei ) ∩ U (ej ) = ∅ whenever i 6= j. Without loss of generality ei ej = 0 whenever
i 6= j. Therefore, all the conditions of Theorem 3.12 are satisfied except for possibly (2).
We claim that V (2) ∩ U (e1 ) and V (a + b) ∪ V (a − b − 1) are disjoint closed sets. To see
this let M ∈ V (2) ∩ U (e1 ). Suppose that M ∈ V (a + b). Since M ∈ V (2), M ∈ V (a − b) and
hence M ∈ U (f2 ), contradicting that M ∈ U (e1 ). If M ∈ V (a − b − 1), then M ∈ V (a + b − 1)
which implies that M ∈ V (f1 ), again contradicting that M ∈ U (e1 ). Similarly, V (2) ∩ U (e2 )
and V (a − b) ∩ V (a + b − 1) are disjoint closed sets.
Since A is clean it follows that the collection of idempotent clopen sets form a base for
the topology. It follows that there is an idempotent clopen subset K1 ⊆ U (e1 ) such that
V (2) ∩ U (e1 ) ⊆ K1 and (V (a + b) ∪ V (a − b − 1)) ∩ K1 = ∅. Similarly, there is an idempotent
clopen subset K2 ⊆ U (e2 ) such that V (2) ∩ U (e2 ) ⊆ K2 and (V (a − b) ∪ V (a + b − 1)) ∩ K2 = ∅.
By combining U (e4 ) together with K1 and K2 and shrinking down U (e1 ) and U (e2 ) in the
appropriate manner, we obtain idempotent clopen subsets which satisfy conditions (1)-(6) of
Theorem 3.12. Therefore, A[Z2 ] is clean.
Corollary 3.14. Suppose G is an elementary 2-group. Then A[G] is clean if and only if A is
clean.
Proof. Since G is an elementary 2-group it follows that for each f ∈ A[G] there exists a
finite subgroup H of G such that f ∈ A[H] and H is a finite direct product of copies of Z2 .
It follows that A[H] is a finite product of copies of A[Z2 ] which is clean. Therefore, f ∈ A[H]
can be written as a s um of a unit and an idempotent in A[H], and hence can be written as a
sum of a unit and an idempotent in A[G].
12
Warren Wm. McGovern
References
[1]
D.D. Anderson, V.P. Camillo, Commutative rings whose elements are a sum of a unit and an
idempotent, Comm. Alg. 30 no. 7 (2002) 3327-3336.
[2]
I. Connell, On the group ring, Can. J. Math. 15 (1963) 650-685
[3]
R. Gilmer, Commutative Semi-Group Rings, Univ. Chicago Press, Chicago, 1984.
[4]
J. Han, W.K. Nicholson, Extensions of clean rings, Comm. Alg. 29 no. 6 (2001) 2589-2595.
[5]
P.T. Johnstone, Stone Spaces, Cambridge Stud. Adv. Math., 3, Cambridge Univ. Press, Cambridge, 1982.
[6]
W.Wm. McGovern, Neat rings, preprint.
[7]
W.K. Nicholson, Lifting idempotents and exchange rings, Trans. A.M.S., 229 (1977) 269-278.
```